Thanks Lars!
What is your (high-level) design process and how much do you rely on listening?
Last edited:
-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
What role should listening play in speaker design
- Thread starter amirm
- Start date
Thanks Lars!
What is your (high-level) design process and how much do you rely on listening?
- Joined
- Sep 16, 2019
- Messages
- 1,418
- Likes
- 3,375
It was around €1000 without VAT five years ago. Expect more now..
a4eaudio
Addicted to Fun and Learning
The answer probably belongs in one of the existing threads on measurements rather than here, but I'd be interested in understanding why.
I heard about $1,200 in the US, probably 2-3 year ago, which is consistent with that.
In the video below between 5:09-7:28 in the interview. About the sonic differences between Ncore and Purifi Eigentakt.
This is what Bruno Putzeys says:
The thereness of the sound..
The sound became finer, warmer. Emotionally more engaging. Softer, gentler. ......More relaxing, more open and warmer.
What do I know, can you hear it? Maybe if the different modules are pushed to the clipping limit together with extremely load tricky, extremely low sensitive speakers (plus that the speaker itself is not the bottleneck) then I can buy it.
Do the Ncore modules have audible distortion in the higher frequencies?
But the old notion that class D is only good for bass tasks hardly applies to the Ncore modules?
Edit:
Bruno Putzeys should be given credit for contradicting that audiophile nonsense about negative feedback having a negative impact. Quite the opposite, the negative is the positive so to speak.
Bruno Putzeys of course puts it much better, from 7:37 in the video.
Last edited:
There have been many explanations around dips in the region; e.g. reducing the problem of driver integration. Another obvious error is the dip at 1.8 kHz followed by a peak 3-4 kHz in stereo setups. It is quite easily detected by listening in a pair of speakers that show perfect linear on-axis measurements.
My design flow is heavily based on model based optimisation before building anything. this includes use of FEA and Matlab. I start with the bass response and radiation pattern since these determine the size and shape of the cabinet. These things are hard to change later in the process. The bass response model combines a driver model with a box model and the filter behind. All these interact, because the bass port tuning affects the driver impedance which again changes the electrical response of the filter. In addition we have the box diffraction (baffle step) that we can simulate accurately using FEA. All this can be optimised numerically to find the optimal cabinet volume, box tuning and woofer filter.
Now working with my new colleague Damian. He builds multiphysics models of the complete box and drivers combining acoustics and solid mechanics. This directs how to brace the cabinet and handle resonances. Also, placement of damping material can be optimised.
Next is to measure the response of each driver and optimise the crossover filter. Here we again use numerical optimisation that targets both the summed response flatness and the phase match between the drivers. In addition we can constrain the impedance of the speaker.
Next is to build and verify the filter and measure again to check the acoustic response.
Listening may show that some compromises has to be changed. Eg we need a more complex filter to flatten the response further.
that was the short story
Maybe those designers are using their ears rather than the optimizer.
But far, far more common than it was a decade or two ago.
a4eaudio
Addicted to Fun and Learning
True, but I could make a list of 25 things that "maybe" are the cause. But, I'm trying to understand why 100% of the commercial speaker designers that have posted in this thread (which I think is now up to 3 people) listen to their speakers when it is supposedly quite easy to design a speaker that measures excellently and will assuredly sound good without listening to confirm it.
kimmosto
Senior Member
- Joined
- Sep 12, 2019
- Messages
- 443
- Likes
- 1,014
That could be okay for final approval and documentation, but probably more expensive for development including extra driver measurements and possible iterations and variants.
I remember (barely) my first designs which were measured in anechoic in Otaniemi Helsinki. Guys measured total on-axis and each four ways separately. Cost was 400 mk in 1986. That would be ca. 155 euros today
Just wonder if anyone is using ACE-bass in active speaker construction nowadays, or if it is no gains to be had?
As risk of putting a *should-bird* into flight... don't want anybody to get *should*
on. lol
......here's what I like as best practice for DIY speaker design, in terms of measurements vs listening..
First, I've never listened to any DIY speaker or sub, until its xovers, EQs, levels, delays, etc... all processing ...is fully in place.
All parameters are determined solely by measurements.
I tune to flat On-axis frequency magnitude response. (And a step further with either flat linear phase, or smooth minimum phase rotation.
Quick spins are also taken to make sure the design has no uncorrectable flaws for when I move on to the next stage of directivity optimization.
Then I get to stop and listen...
About the only thing I don't listen for is tonality...
I know about what to expect given the flat on-ax response.
And more importantly IT DOESN'T MATTER.
Tonality and preference curves are simply not worth the endless consideration & debate, when so easy to adjust a speaker to meet individual preferences and circumstances..
Everyone *should* have tone controls of some sort ....it is plain foolish not to imo... for a host of reasons. Sorry if that offends anyone.
What I do listen for first and foremost is clarity. Then transients, and a sense of rhythm..
How does the speaker hold up when pushed? Especially down low. Conversely, how does it sound at reduced volume.
How does it vary vs angle and vs distance. What does it sound like from the next room. Outdoors... (The best listening test ime)
If it's a keeper, it's time to optimize directivity. Which is the xover fine tuning stage, and must be done via measurements, ime. (Who can reliably assess off-axis behavior by ear/tonality?)
Xovers are not for adjusting frequency response, or voicing, etc. They are for establishing the smoothest directivity possible given the speaker's physical acoustic design.
The physical design of the speaker dictates a rather narrow range of optimal crossovers, and *should* not be tampered with once set by measurements.
If someone want to voice or add preference curves, that's best done as a global overlay on the finished speaker.
Hey, you could call that adding tone controls, huh? !!
Then, when I have two fine measuring on-ax, and good directivity speakers, it's finally stereo listening time.
There is no further speaker processing to do.
It's speaker placement time, room treatment time, maybe whack down a room mode with EQ...
....and of course, tone control time! (Even if it's set once and leave alone. No comments on DSP room correction. )
So, my 2c really...
Why do we have to belabor whether voicings or preference curves *should* go into speaker design ????
I say NO, make speakers flat and let's all get some form of damn tone controls.... "for speakers sake" <sigh>....
on. lol......here's what I like as best practice for DIY speaker design, in terms of measurements vs listening..
First, I've never listened to any DIY speaker or sub, until its xovers, EQs, levels, delays, etc... all processing ...is fully in place.
All parameters are determined solely by measurements.
I tune to flat On-axis frequency magnitude response. (And a step further with either flat linear phase, or smooth minimum phase rotation.
Quick spins are also taken to make sure the design has no uncorrectable flaws for when I move on to the next stage of directivity optimization.
Then I get to stop and listen...
About the only thing I don't listen for is tonality...
I know about what to expect given the flat on-ax response.
And more importantly IT DOESN'T MATTER.
Tonality and preference curves are simply not worth the endless consideration & debate, when so easy to adjust a speaker to meet individual preferences and circumstances..
Everyone *should* have tone controls of some sort ....it is plain foolish not to imo... for a host of reasons. Sorry if that offends anyone.
What I do listen for first and foremost is clarity. Then transients, and a sense of rhythm..
How does the speaker hold up when pushed? Especially down low. Conversely, how does it sound at reduced volume.
How does it vary vs angle and vs distance. What does it sound like from the next room. Outdoors... (The best listening test ime)
If it's a keeper, it's time to optimize directivity. Which is the xover fine tuning stage, and must be done via measurements, ime. (Who can reliably assess off-axis behavior by ear/tonality?)
Xovers are not for adjusting frequency response, or voicing, etc. They are for establishing the smoothest directivity possible given the speaker's physical acoustic design.
The physical design of the speaker dictates a rather narrow range of optimal crossovers, and *should*
not be tampered with once set by measurements.If someone want to voice or add preference curves, that's best done as a global overlay on the finished speaker.
Hey, you could call that adding tone controls, huh? !!
Then, when I have two fine measuring on-ax, and good directivity speakers, it's finally stereo listening time.
There is no further speaker processing to do.
It's speaker placement time, room treatment time, maybe whack down a room mode with EQ...
....and of course, tone control time! (Even if it's set once and leave alone. No comments on DSP room correction. )
So, my 2c really...
Why do we have to belabor whether voicings or preference curves *should*
go into speaker design ????I say NO, make speakers flat and let's all get some form of damn tone controls.... "for speakers sake" <sigh>....
good question. I was fascinated by the method many years back. Carl Erik Ståhl wrote an AES paper way back in the golden days of hifi. The negative output impedance is helping to reduce some distortion mechanisms but also make others worse (kind of the opposite of current drive). Without a distortion reduction then we can just EQ with a Linkwitz Transform.
- Joined
- Sep 16, 2019
- Messages
- 1,418
- Likes
- 3,375
I assume COMSOL which in turn is also used to develop the drivers themselves?
yes Comsol.
Jim Taylor
Major Contributor
- Joined
- Apr 7, 2024
- Messages
- 1,322
- Likes
- 4,468
Gosh, an n=3!
Anyway, I think the sarc in my post maybe escaped you?
Here's an n=1 for ya: I would definitely buy a speaker that had good Klippel performance before I would buy one on just some designer's recommendation. No question.
In fact I never 'audition' speakers before buying them -- auditioning in a retail salon always struck me as silly, given the vast variance room acoustics,, and I hadn't the hobbyist intensity to try-then-return multiple speakers at home.
So, what to do? Rely on reviews and half-assed measurements therein?. A crapshoot, but that's what was available.
But then after years of crapshooting, science came to my rescue. Buying a speaker 'sound unheard' is a far less fraught proposition now than it was when I was a youth. Thanks to good measurements.
a4eaudio
Addicted to Fun and Learning
Feel free to substitute "good" with whatever descriptive adjective you find most useful to further the conversation. To be clear, I have no objection to "accurate".Because "good" and "accurate" are two different things.
Because "good" and "accurate" are two different things.
Is it? And if so, why would anyone want accurate for leisure listening? Surely you'll want your music to sound as good as possible?
a4eaudio
Addicted to Fun and Learning
There is no such thing as "good Klippel performance". The Klippel device will objectively measure the speaker, even if it is horrible. The main thing in this thread that has been proposed as an objective measure related to sound quality is linear/flat/ruler flat On-axis response and I am pretty sure no one has even bothered to define that. Is +/- 2dB or +/- 0.5dB ruler flat? What level of smoothing? Over what frequency range? And if only ruler flat On-axis axis matters, why bother with Spinorama?
To be clear, I'm not just being argumentative. Here is an hypothetical example... ASSUME you can design a speaker with a very flat on-axis response with one xo variation producing a smooth PIR with slope -0.80 and another xo variation with slope -0.90. All of the Spinorama measurements look great and they give nearly identical, high Preference Ratings. How do you decide which one to manufacture?
- Joined
- Sep 16, 2019
- Messages
- 1,418
- Likes
- 3,375
But recordings vary, sometimes a lot. Assuming a very good loudspeaker that's well behaved on- and off-axis, well behaved time domain behaviour (in case Kimmo is reading
) and good dynamic abilities..Even that loudspeaker cannot sound great with every recording. But I'd still rather have that, and some tone controls if need be than something that's designed to not sound offensing, because I assume that is what is being voiced for - making sure nothing sounds (too) bad.
